HBase is simply a non-relational database that runs on HDFS which is essentially a filesystem abstraction layer on top of Hadoop which is map-reduce framework.
Apache HBase provides Bigtable-like capabilities on top of Hadoop and HDFS.
So the breakdown is like this,
HDFS is a distributed file system that is well suited for the storage of large files. ...
Cassandra definitely is more suitable for storage of metadata only, when you have big payloads, it's performance isn't very good. Similar stuff was for HBase when I did use it several years ago.
For storage of binary data itself, I maybe would go with something S3-compatible.